We evaluated prospectively the association of smoking and other potential risk factors with bladder carcinoma incidence in postmenopausal women.
We evaluated prospectively the association of smoking and other potential risk factors with bladder carcinoma incidence in postmenopausal women.
A total of 37,459 women participating in the Iowa Women's Health Study completed baseline questionnaires in 1986 and were followed 13 years for bladder carcinoma incidence (n = 112).
Adjusted for potential confounders, the relative risk (RR) of bladder carcinoma in women who were current smokers compared with those who had never smoked was 3.58 (95% confidence interval [CI] = 1.86–6.88). The RR declined as years since quitting increased. Currently, married women, compared with unmarried women, had a RR of 0.66 (95% CI = 0.44–0.99). A 2.46-fold (95% CI = 1.32–4.59) increase in bladder carcinoma risk was identified for women who reported, versus did not report, diabetes. Regular versus no physical activity (RR = 0.66, 95% CI 0.43–1.01) and body mass index were inversely associated (P = 0.06) with bladder carcinoma incidence.
We confirmed that cigarette smoking is an important risk factor for bladder carcinoma in women; women who had quit smoking had a reduction of risk. We also identified diabetes as a potential risk factor, which may invite more research on its role in the development of urinary bladder carcinoma. Cancer 2002;95:2316–23. © 2002 American Cancer Society.
Each year, nearly 55,000 people in the United States learn that they have urinary bladder carcinoma.1 Risk factors consistently associated with bladder carcinoma are male gender, older age, a history of cigarette smoking, occupational exposure to aromatic amines (e.g., in the dye and rubber industries), and bladder infections. Other suggested risk factors include alcohol and coffee intake, use of some drugs such as phenacetin-containing analgesics, and exposure to ionizing radiation.2–8 Cigarette smoking is the key factor contributing to bladder carcinoma in the United States.9 Among many case–control studies and a few cohort studies,10–25 the relative risk (RR) of bladder carcinoma for smokers versus nonsmokers varied between 1.2 and 6.3.
Many studies have investigated risk factors for bladder carcinoma in men.18–25 Those that studied both men and women were mostly case–control studies that included few women.10–17 Therefore, risk factors for bladder carcinoma in women have received little attention. We investigated cigarette smoking and other potential risk factors for bladder carcinoma in a prospective study of older women.
In January 1986, a mailed questionnaire was completed by 41,836 (42.7%) of 98,030 randomly selected women, between the ages of 55 and 69 years, who had a valid Iowa drivers license in 1985. The baseline questionnaire included standard questions on education level, usual alcohol intake during the past year, hormone replacement therapy use, marital status, diabetes, hypertension, history of blood transfusion, and main lifetime occupation. Participants were asked their smoking status (never, current, past). Former smokers were questioned regarding their age at the time of quitting. Smoking frequency was derived from questions about the number of cigarettes smoked each day. Participants were asked if they were involved in any leisure time exercise, and if so, the frequency of moderate and vigorous activities. Examples of moderate activity were bowling, golf, light sports, gardening, and walking and examples of vigorous activity were jogging, racket sports, swimming, aerobics, and other strenuous sports. The two questions on frequency were combined to create a three-level activity score (low, medium, and high).
Enclosed with the questionnaire were a paper tape measure and written instructions to have a friend measure circumference of the waist (2.5 cm above the umbilicus) and hips (maximal protrusion). The waist-to-hip ratio (WHR) was calculated using these values. Body mass index (BMI) was derived from the current height and weight as reported by the participants. The anthropometric variables obtained by this protocol were valid and reliable.26 Prevalent carcinomas were ascertained by asking the women whether they had ever been told by a physician that they had any form of carcinoma, excluding skin carcinoma. No data were available on family history of bladder carcinoma or bladder carcinoma in male partners.
Follow-up questionnaires were sent in 1987 (91% response), 1989 (89% response), 1992 (83% response), and 1997 (79% response) to update the vital status of the cohort. Deaths among nonrespondents to follow-up questionnaires were found through the National Death Index. Incident cases of bladder carcinoma between 1986 and 1998, along with Iowa deaths, were identified by linking participant identifiers to the State Health Registry of Iowa, which includes a National Cancer Institute-supported Surveillance, Epidemiology, and End Results (SEER) cancer registry.
Of the 41,836 participating women, we excluded from consideration those who were premenopausal (n = 569) and those who reported a baseline history of carcinoma other than skin carcinoma (n = 3830). After excluding 4377 women, 37,459 women remained for the analysis. Cohort members were at risk of developing bladder carcinoma until a registered bladder carcinoma diagnosis, death, and identified date of emigration from Iowa (approximately 0.5% annually) or December 31, 1998, whichever came first. Otherwise, person-years accumulated from the date of the 1986 baseline questionnaire through December 31, 1998. Women were categorized according to various baseline characteristics, using quintiles or natural categories. We computed age-adjusted incidence rates of bladder carcinoma per 100,000 person-years for each category by Poisson regression (SAS program GENMOD). Multivariate-adjusted RRs, 95% confidence intervals (CIs), and interaction tests were computed by proportional hazards regression (SAS program PHREG). In developing the multivariate models, all potential risk factors were entered into the regression model. The factors that were not materially associated with bladder carcinoma were removed, although their removal did not greatly alter the remaining parameter estimates. Tests for trend in RRs were modeled using ordinal categories. The population attributable risk (PAR %) for ever smoking was computed from the RR and prevalence of the exposure (p) using the formula:
Over the 13-year follow-up period, we identified 112 incident cases of urinary bladder carcinoma in the cohort of at-risk women. The incident rate was approximately 25 per 100,000 person-years of observation (Table 1) and increased with age.
|Baseline variable||Prevalence of variables (%)||No. (events)||Person-years||Incidence rate per 105 person-years||Relative risk||95% confidence interval|
|Age (not age-adjusted) (yrs)|
|Quit ≤ 5 yrs||5||10||19,508||62.8||3.76||1.88–7.49a|
|Quit > 5 to 15 yrs||5||7||21,215||39.9||2.37||1.06–5.28a|
|Quit > 15 yrs||9||6||36,745||19.8||1.17||0.50–2.75|
|Pack-years of smoking|
|Regular physical activity|
|Body mass index (kg/m2)|
|Hip circumference (cm)|
|Waist circumference (cm)|
|Age at menarche|
|Hormone replacement therapy|
|Never or < 1/mo||28||28||118,194||22.9||1.00|
|1/mo to 5–6/wk||19||19||79,048||23.2||1.00||0.56–1.79|
|1/day to 2–3/day||34||34||144,297||22.9||1.00||0.60–1.64|
|Never or < 1/mo||42||45||174,719||25.5||1.00|
|1/mo to 5–6/wk||41||51||173,259||28.9||1.13||0.76–1.69|
|1/day to 2–3/day||15||10||64,393||15.2||0.59||0.30–1.18|
|≤ 4 g/day||24||26||105,147||29.2||1.14||0.72–1.81|
|> 4 g/day||20||30||87,456||41.1||1.61||1.03–2.52*|
|1–3 per mo||15||14||63,034||25.7||0.84||0.48–1.47|
|> 1–3 per mo||7||4||27,723||16.7||0.55||0.20–1.49|
|1–3 per mo||13||14||55,783||29.3||0.95||0.54–1.67|
|> 1–3 per mo||9||4||34,950||13.5||0.43||0.16–1.19|
|1–3 per mo||8||5||34,662||17.3||0.59||0.24–1.44|
|> 1–3 per mo||10||14||41,801||41.1||1.39||0.79–2.46|
|1–3 per mo||11||15||46,870||38.5||1.36||0.78–2.37|
|> 1–3 per mo||15||18||62,717||34.2||1.21||0.72–2.02|
|< High school||20||20||86,050||26.4||1.00|
|> High school||39||38||170,970||26.0||0.99||0.57–1.69|
Approximately 15% of the women in the cohort were current smokers and 66% were never-smokers (Table 1). The age-adjusted incidence rates of bladder carcinoma were 5.49-fold higher in current smokers (91.4 per 100,000 person-years) than in nonsmokers (17.1 per 100,000 person-years). After multivariate adjustment for physical activity, body size, diabetes, alcohol consumption, marital status, and occupation, the RR for current versus never-smokers was 4.23 (95% CI = 2.76–6.70). After further adjustment for pack-years of cigarette smoking and time since quitting, the RR was 3.58 (95% CI = 1.86–6.88; Table 2). There was an increase in the RR of bladder carcinoma as the number of pack-years increased (P < 0.001; Table 1). In the multivariate model, each 20 pack-years increase in smoking was associated with a 1.13-fold increase in the RR of bladder carcinoma (95% CI = 0.88–1.43). A potential beneficial effect of smoking cessation was apparent as bladder carcinoma incidence decreased as the time since quitting increased (P < 0.001). The incident rate for those who quit more than 15 years ago approached that of nonsmokers.
|Variablea||Relative risk||95% Confidence interval|
|Age groups (yrs)|
|Pack-years of smoking (20 pack-years)||1.13||0.88–1.43|
|Quit ≤ 5 yrs ago||2.89||1.25–6.66|
|Quit > 5 to 15 yrs ago||1.69||0.67–4.25|
|Quit > 15 yrs ago||1.06||0.44–2.54|
|Regular physical activity|
|Body mass index (kg/m2)|
|≤ 4 g/day||1.08||0.66–1.75|
|> 4 g/day||1.05||0.65–1.70|
Approximately 42% of the cohort reported regular physical activity. These women had a 41% lower age-adjusted incident rate of bladder carcinoma compared with women who were not physically active (Table 1). However, a dose response was not evident using the physical activity index. After multivariate adjustment, the RR of bladder carcinoma for regular physical activity was only slightly attenuated to 0.66 (95% CI = 0.43–1.01; Table 2).
Six percent of the cohort reported diabetes at baseline. They experienced a 1.99-fold greater age-adjusted incidence rate of bladder carcinoma compared with women without diabetes. After multivariate adjustments, the RR was 2.46 (95% CI = 1.32–4.59; Table 2).
Compared with women who were currently unmarried, women currently married had a 34% lower incidence of bladder carcinoma (RR = 0.66, 95% CI = 0.44–0.99), after multivariable adjustments (Table 2).
Compared with never drinking, drinking more than 4 g per day of alcohol was associated with increased bladder carcinoma (age-adjusted RR = 1.61, 95% CI = 1.03–2.52; Table 1). However, as shown in Table 2, after adjustment for other risk factors (chiefly smoking) the RR was 1.05 (95% CI = 0.65–1.70). The type of alcohol consumed was not associated with bladder carcinoma incidence.
Women who described their occupation as farming or crafts related had increased bladder carcinoma incidence (age-adjusted RR = 1.84 95%, CI = 1.11–3.10) compared with homemakers and those who never worked. Professionals and clerical workers also had an increased age-adjusted incidence (RR = 1.51, 95% CI = 0.98–2.31) compared with homemakers and those who never worked. These associations did not remain statistically significant after adjustment for other risk factors (chiefly smoking).
Women in the highest versus women in the lowest quintile of BMI had a reduced age-adjusted RR of 0.53 (95% CI = 0.29–0.96, P = 0.01). On multivariate adjustment, the RR was 0.63 (95% CI = 0.33–1.19, P = 0.06).
None of the other risk factors examined in Table 1 was associated with bladder carcinoma incidence: WHR and other anthropometric variables, educational level, hypertension, history of blood transfusion, coffee or tea intake, age at menarche, parity, or hormone replacement therapy use. The age-adjusted RR for heavy versus never or rare coffee drinking was 1.59 (95% CI = 0.95–2.68, P = 0.25), but this was reduced to 1.39 upon multivariate analysis.
There was no evidence of interaction of smoking with diabetes, physical activity, BMI, or age (data not shown).
In general, women have lower rates of bladder carcinoma than do men,1 in part explained by women's lower exposures to occupational hazards and to smoking. However, published data on bladder carcinoma risk factors in women remain sparse. The main finding of this large prospective study of older women was that cigarette smoking was a strong risk factor for bladder carcinoma. Based on a 34% prevalence of ever smoking and a RR of 2.0, PAR indicated that approximately 25% of bladder carcinoma incident cases in our study population might be attributed to ever smoking. Other risk factors for bladder carcinoma in this study included age, diabetes, and currently being unmarried (vs. being married). Our findings also suggest the possibility of an association of increased bladder carcinoma incidence with reduced physical activity and BMI. There was no independent association with alcohol intake or others such as hypertension or hormone therapy use.
The association between smoking and bladder carcinoma, first identified by Lilienfeld et al. in 1956,27 has been verified by numerous studies.10–25 However, relatively few studies10–17 involved women. Previous RR estimates for cigarette smoking and bladder carcinoma in women ranged from 2.4 to 5.2.10–17 These RR estimates were higher than those for men with comparable numbers of cigarettes smoked. This may be because women are more susceptible to smoking-induced bladder neoplasms or, more likely, because women typically have fewer other bladder carcinoma risk factors than do men. We observed in these postmenopausal women that current smokers had approximately four times the risk of bladder carcinoma of those who never smoked. The risk was related to the number and duration of cigarettes smoked.
Based on existing evidence, the association between smoking and bladder carcinoma has been deemed causal by the International Agency for Research on Cancer2 and is believed to result from carcinogenic aromatic amines in tobacco smoke. The entire mechanism has not been determined. However, laboratory and epidemiologic data indicate that metabolism of arylamines is under genetic control and that the risk of carcinoma is negatively correlated with the ability to detoxify these compounds. Metabolic activation of arylamines results in DNA adduct formation in the bladder mucosa and carcinomas likely arise due to the resultant mutations. The detoxification of arylamines occurs via polymorphic phase 2 enzymes, including N-acetyltransferase and glutathione S-transferase M1.28, 29
Although not a randomized trial, the reduction of bladder carcinoma risk with cessation of smoking additionally supports the association being causal. In women who had quit smoking, the incidence declined as the years since quitting increased. After more than 15 years since quitting, incidence approached that of never smoking. The risk reduction appeared to be quite substantial within the first 5 years of quitting, compared with current smoking, suggesting that smoking may have a late effect on carcinogenesis, as supported by some studies.9, 12, 20
Few studies have evaluated the association of diabetes with bladder carcinoma. We observed that women with diabetes had a 2.46-fold greater risk of bladder carcinoma than women without diabetes, after taking other risk factors into consideration. Risk of bladder carcinoma remained statistically significantly elevated in women with diabetes, even after excluding women who had developed bladder carcinoma within the first year of follow-up (data not shown). This lessens the possibility of reverse causality, namely, that the stress of preclinical carcinoma led to greater manifestation of diabetic symptoms. Kantor et al.7 reported a statistically significant RR of 1.21 for the association between diabetes and bladder carcinoma Ragozzino et al.30 reported the RR to be 1.5 (95% CI = 0.6–3.1). Several other studies found no association.31–35
Several possibilities might explain an association for diabetes. First, we identified diabetes by self-report and therefore some women could have been misclassified on diabetes status. However, others have reported reasonable validity of diabetes self-report.36 Typically, misclassification of exposure would lead to underestimated, not overestimated, associations. Second, women with diabetes may have had more urine examinations and had carcinoma detected earlier, thereby leading to a detection bias. Alternatively, diabetes may truly be a risk factor for development of urinary bladder carcinoma because it is associated with a range of dietary, immunologic, metabolic, and hormonal abnormalities. For example, decreased insulin sensitivity and the resultant increased circulating insulin may promote tumor growth. Diabetes leads to increased risk of urinary tract infection,37 which is an established risk factor for bladder carcinoma2, 9 and neurogenic bladder abnormalities.38 It is possible that the few previous studies that evaluated diabetes and bladder carcinoma did not have adequate statistical power to detect the association.
We found a lower risk of bladder carcinoma in married women. The literature indicates that married persons are at lower risk for all carcinomas, including urinary bladder carcinoma, compared with those who are single, divorced, or widowed.39–41 Ernster et al.40 reported lower standardized incidence ratios for all carcinomas combined, as well as for bladder carcinoma, in married versus not married women aged 55 and older. Better health habits of married people41 might explain the findings, although they did not in our study.
In accordance with many previous studies, mostly comprising men, our results do not suggest an association between either the quantity or type of alcohol consumed and bladder carcinoma.8, 42, 43 Likewise, finding no association of bladder carcinoma with coffee consumption is consistent with most previous studies.2, 6, 16, 17, 19
Several lifestyle factors have not been evaluated in relation to bladder carcinoma. Women who engaged in regular physical activity showed a decreased risk of bladder carcinoma compared with women who did not. Overweight women had a decreased risk of bladder carcinoma compared with women in the lowest BMI category. However, these two factors were not quite statistically significant after adjusting for smoking. It is possible that increased physical activity and increased BMI truly do reduce the risk of bladder carcinoma, but the number of events was too small to confirm this. For example, physical activity might reduce cancer risk because it can improve insulin sensitivity and reduce circulating insulin levels. The inverse association for BMI most likely is due to residual confounding by smoking (i.e., thinner women were heavier smokers and this was not fully accounted for by statistical adjustment).
Limitations of our study included relatively few cases and missing information on some bladder carcinoma risk factors, such as fluid intake, history of urinary tract infections, and specific occupational exposures. The RR estimates had fairly wide CIs and some confounding bias may not have been entirely controlled.
To conclude, we confirmed that cigarette smoking is an important risk factor for bladder carcinoma in women; women who had quit smoking had a reduction of risk. We also identified diabetes and lifestyle factors such as reduced physical activity and BMI as potential risk factors, which may invite more research on their role in the development of urinary bladder carcinoma.